Re: [PATCH][RFC] Proposal for a new watchdog interface using sysfs

[Alan Cox]

On Fri, 2003-02-14 at 13:48, [e-mail] wrote:
> I admit not having thoroughly read the source to check - is the userspace API
> for accessing  all these chips fairly uniform and rational, so that a user
> program can be reasonably sure that if stat("/dev/watchdog") returns zero, that
> it knows how to deal with it?  Or are they all sufficiently close to the "keep
> reloading a countdown timer from userspace, and if it ever doesn't get reloaded,
> kick the kernel in the seat of the pants" programming model?  Of course, even
> a disagreement on the units of the timer could be bad - a seconds/milliseconds
> clash could result is a *real* fast lack-of-joy situation.. ;)

watchdog interfaces have a defined API, which they all follow fairly closely. That
makes adding watchdogs as a device class nice and easy